Development and Validation of a Risk Prediction Model for Ketosis-Prone Type 2 Diabetes Mellitus Among Patients Newly Diagnosed with Type 2 Diabetes Mellitus in China

BACKGROUND: We established a nomogram for ketosis-prone type 2 diabetes mellitus (KP-T2DM) in the Chinese adult population in order to identify high-risk groups early and intervene in the disease progression in a timely manner. METHODS: We reviewed the medical records of 924 adults with newly diagnosed T2DM from January 2018 to June 2021. All patients were randomly divided into the training and validation sets at a ratio of 7:3. The least absolute shrinkage and selection operator regression analysis method was used to screen the predictors of the training set, and the multivariable logistic regression analysis was used to establish the nomogram prediction model. We verified the prediction model using the receiver operating characteristic (ROC) curve, judged the model’s goodness-of-fit using the Hosmer-Lemeshow goodness-of-fit test, and predicted the risk of ketosis using the decision curve analysis. RESULTS: A total of 21 variables were analyzed, and four predictors—hemoglobin A1C, 2-hour postprandial blood glucose, 2-hour postprandial C-peptide, and age—were established. The area under the ROC curve for the training and validation sets were 0.8172 and 0.8084, respectively. The Hosmer-Lemeshow test showed that the prediction model and validation set have a high degree of fit. The decision curve analysis curve showed that the nomogram had better clinical applicability when the threshold probability of the patients was 0.03–0.79. CONCLUSION: The nomogram based on hemoglobin A1C, 2-hour postprandial blood glucose, 2-hour postprandial C-peptide, and age has good performance and can serve as a favorable tool for clinicians to predict KP-T2DM.